Raising and ControlIt seems to work like this. An anaphor must have an antecedent within some domain,
say the clause, and cannot refer to anything outside of this, or indeed have no
antecedent at all. A pronominal, on the other hand, cannot have its antecedent within
the same domain, but may refer freely outside this domain. The situation is a little
more complex than this however, as there are restrictions on where the antecedent
must be in relation to the pronoun. For example, while an object anaphor can take the
subject as its antecedent, a subject anaphor cannot take the object as its antecedent:
(75) a the doctor healed himself
b *himself healed the doctor
In general, the antecedent has to be structurally above the pronoun. Let us call a
structurally superior antecedent a binder. The principles involved in determining the
distribution of pronouns can be stated as follows:
(76) a An anaphor must have a binder within the binding domain
b A pronominal cannot have a binder within the binding domain
This is a simplification and things are more complex than this and further
complications can be found if we try to define precisely what the binding domain is.
However, for our purposes it will suffice to know that there is a domain within which
an anaphor must have a binder and a pronominal cannot.
What kind of a pronoun is PRO: an anaphor or a pronominal? Interestingly, this is
not such a straightforward question to answer as PRO demonstrates properties of both
pronominals and anaphors and a number of properties that are unique to itself. One can
find in the literature claims that PRO is a pronominal, an anaphor or even both! One
observation is that PRO more than often must have an antecedent, a property that it
shares with anaphors. Thus, in the following PRO must be taken as referentially
dependent on the subject of the main clause and cannot be taken as referring to
someone not mentioned:
(77) Eddy expects [CP PRO to arrive at noon]
In other contexts, however PRO can lack an antecedent altogether. In these cases, it
gets what is called arbitrary reference, a kind of generic reference similar to that of
the pronoun one:
(78) [CP PRO to be] or [CP PRO not to be], that is the question
While this is similar to the behaviour of a pronominal, it is not entirely equivalent.
For one thing when a pronominal has no antecedent it still has a specific referent
determined by the discourse conditions (i.e. who or what is the topic of the conversation
or who or what is being indicated as the referent in non-linguistic ways such as
pointing, etc.). Moreover, pronominals are free to lack antecedents in general whereas
PRO can only lack an antecedent under specific circumstances. Basically, PRO can
lack an antecedent, and therefore have arbitrary reference when the sentence that PRO
is the subject of is itself a subject or when it is the complement of certain predicates:
(79) a [IP [CP PRO to leave now] would be rude]
b it was not known [CP how [IP PRO to solve the problem]]
c the audience were hard [CP PRO to satisfy]